名古屋工業大学学術機関リポジトリ Nagoya Institute of Technology Repository

Formation and size control of a Ni cluster by

plasma gas condensation

著者（英） Takehiko Hihara, Kenji Sumiyama

journal or JOURNAL OF APPLIED PHYSICS
publication title
volume 84
number 9
page range 5270-5276
year 1998-11-01
URL http://id.nii.ac.jp/1476/00004594/

Copyright (1998) American Institute of Physics. This article may be downloaded for personal use only.
Any other use requires prior permission of the author and the American Institute of Physics.The
following article appeared in Journal of Applied Physics, 84(9), pp.5270- 5276 ; 1998 and may be found
at http://link.aip.org/link/?jap/84/5270
 JOURNAL OF APPLIED PHYSICS VOLUME 84, NUMBER 9 1 NOVEMBER 1998

Formation and size control of a Ni cluster by plasma gas condensation

Takehiko Hiharaa) and Kenji Sumiyama
Institute for Materials Research, Tohoku University, Sendai 980-8577 Japan,
and CREST, Japan Science and Technology Corporation (JST) Japan
~Received 13 May 1998; accepted for publication 24 July 1998!
We have constructed a plasma-gas-condensation type cluster deposition apparatus and tried to find
the optimum operation conditions for controlling the cluster size. Transmission electron microscope
~TEM! observation has been done to evaluate sizes of Ni clusters produced when varying the
volume of a cluster growth region, sputtering power, and inert gas pressure. The mean cluster size
decreases by decreasing the volume of growth region and the sputtering power. The smallest cluster
obtained in this work is about 2.3 nm in diameter. We have considered the following two models for
the cluster growth: ~1! a cluster–cluster collision growth and ~2! an atomic vapor condensation
growth. The cluster growth speed estimated from the former is too slow, while that from the latter
is reasonable in comparison with the present experiments. When stable embryos are made from
atom collisions, they grow up faster and the final cluster sizes estimated from the latter model are
consistent with those observed by TEM. © 1998 American Institute of Physics.
@S0021-8979~98!01321-8#

I. INTRODUCTION growth. Molecular dynamics and Monte Carlo simulations

Studies on nanometric metal and semiconductor clusters
have been motivated by their unique physical and chemical
properties ~different from those of the bulk solids!. Several
experimental techniques have been applied to generate and
characterize small particles and clusters.1–3 Improvement
and development of these cluster sources have been strongly
desired for narrower cluster size distribution and more inten-
sive cluster beam. Accordingly, we need a universal concept
to realize the cluster formation process. A plasma-gas-
condensation ~PGC! type cluster source,4–8 which is a com-
bination of sputtering vaporization and inert gas condensa-
tion techniques, is one of the candidates for this purpose. It is
versatile enough to vaporize transition metals and refractive
elements and is able to control the cluster size by adjusting
the sputter yield, the gas pressure, and the volume of the
cluster growth region.
In previous reports,7,8 we described nanometric Cr and
Co clusters prepared by the PGC type cluster source and the
effects of the sorts, pressure, and the flow rate of inert gases
on the cluster formation. As the Ar gas flow rate increases,
larger-sized clusters are produced. He gas is important to
reduce the cluster size and its dispersion. The mean cluster
size was between 6 and 13 nm.
In order to obtain smaller-sized clusters and to optimize
the operation conditions of PGC, we summarize a guideline
of fine particle formation by gas condensation.9 At the initial
stage of cluster nucleation, a three-body collision is neces-
sary in the gas phase to release the latent heat of condensa-
tion, DE lh , for conserving the energy and momentum of the
total system. When the cluster becomes large enough to
transfer DE lh to its various internal degrees of freedom, two-
body collisions become the main channels for the cluster
a!
Electronic mail: Hihara@snap8.imr.tohoku.ac.jp
have indicated that this transition is expected to occur at the
cluster containing about 7 atoms,10 which is called an em-
bryo. Once the embryo is stabilized, the cluster may further
grow via the following procedures proposed within kinetic
theory of the ideal gas:11,12 a cluster–cluster collision and an
atomic vapor condensation.
In this context, we have constructed another PGC type
cluster beam deposition apparatus, in which the volume of a
cluster growth chamber is variable. This article deals with
the size variation of Ni clusters produced by the PGC appa-
ratus when changing the operational parameters such as sput-
tering power, pressure in the growth chamber, and its vol-
ume. We then discuss the cluster growth mechanism in terms
of the two models mentioned above.
II. PLASMA-GAS-CONDENSATION CLUSTER
SOURCE
A. Overview
Figure 1 shows a schematic diagram of the present PGC
apparatus. The metal vapors are generated by dc magnetron
sputtering. A continuous Ar gas stream adjusted by a mass
flow controller is injected in front of a sputtering target.
Clusters nucleate in a high pressure Ar gas atmosphere,
P Ar50.1– 0.4 kPa, and grow up in the space between the
target, TA, and the nozzle, N1 ~the growth region!. The
length between TA and N1, L g , can vary by moving the
sputtering source back and forth. He gas is also introduced
into the sputtering chamber from the backside of the source
to obtain smaller clusters.8 The cluster beam is extracted
through the N1 nozzle ~5 mm in diameter! by differential
pumping, where we used a powerful mechanical buster pump
whose exhausting speed is 1200 m3/h under the present con-
dition. The cluster beam is further collimated by the three
skimmers, N2, N3, and N4, whose diameters are 5 mm. The

0021-8979/98/84(9)/5270/7/$15.00 5270 © 1998 American Institute of Physics

Downloaded 18 Aug 2010 to 133.68.192.98. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
 J. Appl. Phys., Vol. 84, No. 9, 1 November 1998 T. Hihara and K. Sumiyama 5271

FIG. 1. Schematic diagram of plasma-gas-condensation ~PGC! type cluster deposition apparatus. CM1, CM2: capacitance manometers. TMP: turbomolecular
pump. MBP: mechanical buster pomp. TA: sputtering target.

background pressure of 431027 Pa is attained in the depo-
sition chamber by three turbomolecular pumps. During op-
eration, the pressure of the deposition chamber is typically
231025 Pa.
The microgrid, which was a carbon-coated colodion film
supported by a Cu grid, was used as a substrate for transmis-
sion electron microscope ~TEM! observation. The Ni clusters
were deposited at room temperature. The deposition rate was
measured by a quartz thickness monitor. The samples were
transported through air and TEM observation was performed
with a Hitachi HF-2000 electron microscope, operating at
200 kV.
and that of He, P He5150 Pa at V He54.531024 mol/s ~see
Fig. 2!, for instance, we obtain ṁ51.931024 mol/s for Ar
and 6.931024 mol/s for He gas from Eq. ~1! with T 1
5300 K and A n 52.031025 m2. The estimated ṁ values are
roughly in agreement with V Ar and V He , suggesting that the
gas dynamics through the nozzle is isentropic. The drift ve-
locity of the gas and cluster in the growth region, n z , can be
estimated to be about 0.4 m/s by the following equation:

B. Gas dynamics
The pressure in the growth region, P 1 , and the next
room after the N1 nozzle, P 2 , are measured by capacitance
manometers, CM1 and CM2 as indicated in Fig. 1. Figure 2
shows P 1 and P 2 as a function of the total gas flow rate, V T ,
with the different Ar and He gas flow rates V Ar and V He ,
during the operation.
The mass flow rate, ṁ, through the nozzle can be for-
mulated by the following equation within the standard gas
dynamics,13 because the backpressure, P 2 , is much smaller
than the critical pressure, P * 5 P 1 @ 2/( g 11) # g /( g 21) :

ṁ5
AnP1
Ak B T 1
A S D
gm
2
g 11
~ g 11 ! / ~ g 21 !
, ~1!

where k B is the Boltzmann constant, T 1 is the temperature in

the growth region, A n is the cross section of the N1 nozzle,
and g 55/3 is the ratio of molar heat capacity for the mono- FIG. 2. Inert gas pressure in cluster growth region, P 1 , and the first differ-
atomic He and Ar gases. When the partial pressure of Ar in ential pumping chamber after nozzle, P 2 , as a function of total gas flow,
the growth region, P Ar5130 Pa at V Ar51.531024 mol/s V T 5V Ar1V He .

Downloaded 18 Aug 2010 to 133.68.192.98. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
 5272 J. Appl. Phys., Vol. 84, No. 9, 1 November 1998
FIG. 3. Typical I – V curves of sputtering source and cluster deposition rate
with different Ar gas flows, V Ar , and constant He gas flow.
ṁk B T 1
n z5 , ~2!
mA g P 1
where A g is the cross section of cluster growth region (1.4
31022 m2).
III. RESULTS
Figure 3 shows the dc voltage, V, applied on the Ni
target ~in the left axis! and a deposition rate of clusters on the
substrate ~in the right axis! as a function of the electrical
current, I, between the target and the ground. The voltage is
only moderately sensitive to the current between 0.4 and 0.8
A for V Ar57.431025 – 1.531024 mol/s. This is a typical
behavior for magnetron sputtering. Magnetic fields formed
beyond the target induces a magnetron motion of electrons in
plasma, accelerating ionization of Ar gas above the critical
voltage. This voltage plateau disappears for V Ar.3.0
31024 mol/s. The mean free path of electrons in the plasma
becomes short at high P Ar , so that the ionization efficiency
of Ar is not enhanced further.
With the low Ar flow, V Ar57.431025 mol/s, the depo-
sition rate is one order lower than the others. The embryo
formation and cluster growth are suppressed because the
pressure in the growth region is not enough to thermalize Ni
vapor. The clusters can be observed for V Ar.1.531024 and
V He.3.031024 mol/s as shown in Fig. 4. The cluster size
increases with increasing electrical current, I, as shown in the
insets of Fig. 4. The average cluster size is 3.6 nm for I
51.0 A, while it is 2.3 nm for I50.66 A, where we inves-
tigated the deposition area of 0.016 mm2.
Figure 5 shows the cluster deposition rate as a function
of the growth region length, L g , at V Ar51.531024 and
Downloaded 18 Aug 2010 to 133.68.192.98. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
T. Hihara and K. Sumiyama
FIG. 4. Nanometer size Ni clusters observed by TEM obtained with V Ar
51.531024 and V He53.031024 mol/s. I is current for the sputtering
source and D av is an average cluster size.
V He50 – 7.431024 mol/s. A larger amount of clusters is
produced for longer L g and larger V He . It is obvious that the
He gas flow effectively carries the clusters because the vis-
cosity of He gas is lower than that of Ar gas. Figure 6 shows
the TEM image sequence of Ni clusters prepared with L g
5100, 120, and 170 mm at V Ar51.531024 and V He54.5
31024 mol/s. The corresponding size distributions esti-
mated within the area of 0.034 mm2 are shown in Fig. 7. The
cluster size decreases with decreasing L g because the total
amount of collisions between cluster–cluster and/or cluster–
vapor atom would be proportional to the drift time, L g / n z .
Figure 8 shows the deposition rate versus L g curves for
V Ar53.031024 and V He50 – 7.431024 mol/s. These
curves are totally different from those for V Ar51.5
31024 mol/s. The cluster deposition is detectable even at
L g 560 mm. The deposition rate first increases with increas-
ing L g below 100 mm, but shows two peaks at around L g
5100 and 180 mm. It is also notable that the deposition
monitor markedly fluctuated at L g 5100 and 180 mm. The
 J. Appl. Phys., Vol. 84, No. 9, 1 November 1998 T. Hihara and K. Sumiyama 5273

FIG. 5. Cluster deposition rate as a function of growth length, L g , for

V Ar51.531024 , V He50;7.431024 mol/s, and I51 A.

TEM images inserted in Fig. 8 show a broader distribution in

the cluster size for L g 5180 mm. On the other hand, we ob-
tained monodispersive Ni clusters for L g 580 and 140 mm,
where the deposition rate was very stable. The deposition
rate fluctuation and the broad size distribution at L g 5100
and 180 mm suggest that vortical flows may be induced for
the corresponding growth length and clusters cannot be
transported smoothly.
FIG. 6. TEM images of the samples prepared with varying growth length,
L g 5100, 120, and 170 mm, for V Ar51.531024 , V He54.531024 mol/s,
IV. DISCUSSION and I51 A.

A. Embryo formation
In the PGC process, vaporized metal atoms from the
target rapidly lose their kinetic energy by collisions with the
inert gas atoms. The first step of the cluster growth is embryo
formation by thermodynamical fluctuation through three-
body collisions. This process is very difficult to analyze by a
conventional collision dynamics because there are a lot of
unknown factors and lack of theory. These embryos are so
small that they easily evaporate instead of growing to larger
clusters.14
The cluster formation strongly depends on the presence
of nucleation centers such as impurity atoms, ions, etc., i.e.,
heterogeneous nucleation. In the PGC process, a large
amount of the metal atoms are ionized by penning
process,5,15 where the electrons in the glow discharge region
can excite Ar atoms to the metastable state of 11.5 eV,
whose lifetime of 1 s is much longer than the inverse of the
collision frequency. This energy level is higher than the first
ionization energy of Ni atom ~7.63 eV!,16 so that the fraction
of ionized Ni atom increases, which may promote the em-
bryo formation. This is a dynamic process with slightly pre-
dominating condensation. In this report, however, we neglect
this process and regard simply homogeneous nucleation as a
first approach to the cluster growth.
When we consider the cluster growth, there are two ma-
jor processes. ~1! A cluster–cluster collision process: clusters
react with each other, and ~2! an atomic condensation pro-
cess: the clusters grow with accepting atoms arrived on their
surface one by one. In Secs. IV B and IV C, we try to inter-
pret our experimental results using these concepts.
B. Cluster–cluster collision
In order to estimate the cluster growth, we need a metal
vapor density. Based upon experimental and theoretical re-
sults on sputtering,17,18 the sputtering yield, S y 5N s /N i , is
about 5 for our operating condition, where N i is the number
of incident ions onto the target and N s that of evaporated
atoms. The density of sputtered atoms from the target ~80

Downloaded 18 Aug 2010 to 133.68.192.98. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
 5274 J. Appl. Phys., Vol. 84, No. 9, 1 November 1998 T. Hihara and K. Sumiyama

FIG. 8. Cluster deposition rate as a function of growth length, L g , for

V Ar51.531024 , V He50 – 7.431024 mol/s, and I51 A.

dn k
5 (
C nn2
d t ~ t ! i1 j5k i, j i j (i C i,k n i n k , ~4!
i< j

with

S D
FIG. 7. Size distributions estimated from Fig. 6. Broken lines are the results
1/2
of simulation based on cluster–cluster collisions with frequency t in Eq. ~5!. 8k B T
C i, j 5 p ~ r i 1r j ! 2 .
L g is the growth length and D av is an average cluster size. pm
*
Here, n k is the number of the k-size clusters per unit volume
and the coefficient C i, j is a product of the collision cross
section by the mean speed between the i- and the j-size clus-
mm in diameter! is then about 1.631019 atoms/m3 at I ters. r i 5r 1 i 1/3 is the radius of the i-size cluster composed
51.0 A, provided that there are only Ar ions and electrons in of i atoms and m 5m i m j /(m i 1m j ) is the reduced mass,
*
the initial plasma. where m i and m j are the masses of the i- and j-size clusters.
The supersaturation factor, S, is related with the cluster The initial condition of n k is 1.631019 for k51, as we esti-
radius, r, as follows:19 mated in the present experimental condition, and n k 50 for
kÞ1.
p r 2GV a This approach is based on the view that the growth rate
ln S5ln 5 . ~3! for the k-size cluster is the total sum of the increasing num-
p ` k B Tr
ber of collisions between i- and j-size clusters (k5i1 j) and
Here, G is the surface tension and V a is the atomic volume. the decreasing number of the k-size cluster to larger size
P r and P ` are the equilibrium vapor pressure of the cluster cluster by collisions. The parameter t (t) is the collision
with radius r and that of the plane surface, respectively. frequency11,19
The model applied for the cluster–cluster collision is
simulated by assuming that the cluster grows when it collides
with another, and that the total mass in the system is
t~ t !5 E 0
t
& sn th
n~ t !
V
dt 8 , ~5!

conserved.11 We also neglect the evaporation from the clus-
ter surface, because for large clusters, P r becomes compa-
rable to P ` . The kinetic equation for the collision process is
given by the following equation similar to that for the gas
atom collision:11
where s is the cross section, n th is the thermal velocity of the
clusters, and V is the volume of system. Note that s becomes
larger when a larger cluster meets smaller clusters or vice
versa, indicating that the smaller clusters disappear as soon
as the larger ones are formed.

Downloaded 18 Aug 2010 to 133.68.192.98. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
 J. Appl. Phys., Vol. 84, No. 9, 1 November 1998 T. Hihara and K. Sumiyama 5275

The results of this simulation are shown in Fig. 7 by the

broken lines and overlap well with the size distribution esti-
mated from the TEM results. Here, we obtained t 5120 for
the average cluster size of 3.1 nm. This value corresponds to
t545 s in Eq. ~5!, which is two orders longer than the real
traveling time of 0.25 s for L g 5100 mm. The cluster density
decreases rapidly as clusters grow and it takes a much longer
time for the cluster–cluster collision. Therefore, this colli-
sion process is not appropriate for the present results.

C. Vapor condensation growth

We presume here that some embryos in the growth re-
gion grow into larger clusters by a two-body collision, as
mentioned in Sec. I. The growth rate by adding only atoms
can then be expressed as12
di di
5nz 5 s a,i n thn 0 ~ 12 h ! , ~6!
dt dz
where i is the cluster size, n z is the drift velocity of the
materials as we estimated from the flow rate, and n 0 is the
vapor density. h is a reevaporation to condensation rate fac- FIG. 9. Cluster size, i, dependence on its traveling length, z, obtained from
tor, which consider the evaporation from the cluster surface. numerical solution of differential equation ~9!. n 0 is a vapor density of metal
Note that h can be close to 1 if the cluster is small, i.e., the atoms and D is a fluctuation value of the initial cluster size, i 80 571D.
reevaporation rate, f r , is comparable to the condensation
rate, f c , because the latent heat of condensation of metals is
large, e.g., 4.4 eV for Ni at 0 K and 1 atm. n th is the mean simulation.10 Normalizing Eq. ~8! by the initial pressure ratio
velocity with the reduced mass of the i-size clusters, m i , and of the i 0 -size cluster to vaporized metal atoms, we obtain the
the vapor atom, m a , m 5m i m a /(m i 1m a ), and s a,i is the following equation:
*

S D
collision cross section between them19
di p r 2a i 2/3n 0 8k B T 1/2

n th5 S D
8k B T
pm
*
1/2
, s a,i 5 p ~ r a 1r i ! 2 , ~7! dz
5
nz pma
$ 12exp@ 2GV a u / ~ k B Tr a !# %
~9!
where r a and r i are the radii of the vapor atom and the i-size with
cluster.
u 5 ~ i 1/3
0 2i ! /i 0 i ,
1/3 1/3 1/3
A practical problem to solve the differential equation of
Eq. ~6! is the estimation of h. A heat flow between the clus- where m a is the mass of the vaporized metal atom. The fol-
ter and vapor has been discussed by Knauer12 for the case of lowing approximations are also adopted:
an adiabatic expansion. In his argument, the rate of heat de-
1 1 1 1
livery from vapor to cluster and from cluster to vapor should 5 1 ' ; r i 1r a 'r i 5r a i 1/3.
~10!
be equal during the cluster growth. For the PGC process, m ma mi ma
*
however, the latent heat can be removed by collisions with This differential equation is unstable at the start point,
high dense inert gas atoms, i.e., the clusters are not adiabati- i5i 0 , because f r 5 f c , i.e., the embryo can never grow to a
cally cooled down. A simple explanation for h can be pos- larger cluster. The cluster can, however, become larger by
sible because vaporization is generally proportional to the arriving adatom on the surface if the numerical solution is
vapor pressure:19 f r is proportional to the vapor pressure started from an i 80 value which is slightly larger than the
above a cluster surface with radius r, P r , @see Eq. ~3!# and embryo size, i.e., i 80 5i 0 1D, where 0,D!1.
f c is commensurate to that of material atoms around the The solutions of Eq. ~8! as a function of the traveling
clusters, P n length, z, corresponding to L g in the present experiment are

h}
fr P`
5
fc Pn
exp
2GV a
k B Tr iS D
. ~8!
shown in Fig. 9 with different D values for i 0 57 ~r a
51.24 Å, M a 59.7310226 kg, G51.7 N/m, T5300 K!.
The experimental results obtained in the present work ~see
The main assumption here is that the density of vapor- Figs. 4 and 6! are also indicated. The experimental cluster
ized metal atoms does not decrease with the cluster growth, size at different L g ’s is overlapped well with the solution
i.e., the density of the cluster nucleus is low enough, so that lines estimated for n 0 5831018. The reduction in their size
they do not markedly consume the vaporized metal atoms for by decreasing the sputtering power is also consistent with the
their growth. Equation ~6! can then be solved numerically by numerical solutions for smaller n 0 values ~6.431018 and
setting the initial cluster size, i 0 . As we mentioned before, 5.331018!. One can see that, when D510214, an embryo
the embryo size would be about 7 by Monte Carlo cannot grow. If D is larger than 10214, however, it grows and

Downloaded 18 Aug 2010 to 133.68.192.98. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
 5276 J. Appl. Phys., Vol. 84, No. 9, 1 November 1998
the variation of D value does not conspicuously influence the
final cluster size. This result indicates that clusters practically
grow from embryos. When the embryo is stabilized by
chance, the condensation starts: the cluster growth rate in-
creases by increasing their size because the evaporation from
the cluster surface rapidly decreases.
IV. CONCLUSION
We have constructed a plasma-gas-condensation type
cluster deposition apparatus. The small clusters of about 2
nm in diameter were obtained with the low sputtering power
and the small volume of the cluster growth region. We ap-
plied two models for the cluster growth; cluster–cluster col-
lision and vapor condensation processes. In the cluster–
cluster collision process, the collision frequency decreases as
the cluster size becomes larger. The clusters cannot grow as
fast as the experimental results, even though the size distri-
butions observed by TEM are well fitted by the cluster–
cluster collision simulation. The atomic vapor condensation
process, on the other hand, gives a reasonable explanation
for the real cluster growth. The atoms evaporate easily from
the embryo surface. However, it can grow to a larger cluster,
when its free energy is decreased accidentally by fluctuation.
The final size of the clusters are as large as that obtained by
TEM observations.
ACKNOWLEDGMENTS
The authors wish to thank Dr. T. J. Konno for the sup-
port of the TEM observation, Dr. M. Sakurai for helpful
suggestions, and Dr. S. Yamamuro for useful comments.
They gratefully acknowledge technical contributions to the
PGC cluster deposition apparatus from Vieetech Japan Co.,
Downloaded 18 Aug 2010 to 133.68.192.98. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
T. Hihara and K. Sumiyama
Ltd. They are also indebted to the Laboratory for Develop-
mental Research of Advanced Materials, Institute for Mate-
rials Research, Tohoku University. This work was supported
by CREST ~Core Research for Evolutional Science and
Technology! of Japan Science and Technology Corporation
~JST!.
1
W. A. de Heer, Rev. Mod. Phys. 65, 611 ~1993!.
2
P. Melinon et al., Int. J. Mod. Phys. B 9, 339 ~1995!.
3
H. Haberland, in Clusters of Atoms and Molecules I, edited by H. Haber-
land ~Springer, Berlin, 1995!, p. 207.
4
S. Yatsuya, T. Kamakura, K. Yamauchi, and K. Mihama, Jpn. J. Appl.
Phys., Part 2 25, L42 ~1986!.
5
H. Haberland, M. Karrais, M. Mall, and Y. Thurner, J. Vac. Sci. Technol.
A 10, 3266 ~1992!.
6
G. Hohl, T. Hihara, M. Sakurai, T. Oishi, K. Wakoh, K. Sumiyama, and
K. Suzuki, Jpn. J. Appl. Phys., Part 1 33, 1509 ~1994!.
7
S. Yamamuro, M. Sakurai, T. J. Konno, K. Sumiyama, and K. Suzuki, in
Similarities and Differences Between Atomic Nuclei and Clusters, edited
by Y. Abe et al. @AIP Conf. Proc. 416, 491 ~1997!#.
8
S. Yamamuro, K. Sumiyama, M. Sakurai, and K. Suzuki, Supramol. Sci.
~in press!.
9
C. G. Granqvist and R. A. Buhrman, J. Appl. Phys. 47, 2200 ~1976!.
10
M. Kappes and S. Leutwyler, in Molecular Beams of Clusters in Atomic
an Molecular Beam Methods, edited by G. Scoles ~Oxford University
Press, New York, 1988!, Vol. 1, p. 380.
11
J. M. Soler, N. Garcı́a, O. Echt, K. Sattler, and E. Recknagel, Phys. Rev.
Lett. 49, 1857 ~1982!.
12
W. Knauer, J. Appl. Phys. 62, 841 ~1987!.
13
O. F. Hagena, Surf. Sci. 106, 101 ~1981!.
14
A. Milchev and J. Malinowski, Surf. Sci. 156, 36 ~1985!.
15
B. Chapman, Glow Discharge Processes ~Wiley, New York, 1982!.
16
H. Haken and H. Wolf, The Physics of Atoms and Quanta ~Springer, New
York, 1996!.
17
J. L. Vossen and W. Kern, Thin Film Processes ~Academic, New York,
1978!.
18
P. Sigmund, Phys. Rev. Lett. 184, 383 ~1969!.
19
P. W. Atkins, Physical Chemistry ~Oxford University Press, New York,
1995!.